system-linux.c

   1 /*
   2  * netifd - network interface daemon
   3  * Copyright (C) 2012 Felix Fietkau <nbd@openwrt.org>
   4  * Copyright (C) 2013 Jo-Philipp Wich <jow@openwrt.org>
   5  * Copyright (C) 2013 Steven Barth <steven@midlink.org>
   6  * Copyright (C) 2014 Gioacchino Mazzurco <gio@eigenlab.org>
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License version 2
  10  * as published by the Free Software Foundation
  11  *
  12  * This program is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  * GNU General Public License for more details.
  16  */
  17 #define _GNU_SOURCE
  18
  19 #include <sys/socket.h>
  20 #include <sys/ioctl.h>
  21 #include <sys/stat.h>
  22 #include <sys/syscall.h>
  23
  24 #include <net/if.h>
  25 #include <net/if_arp.h>
  26
  27 #include <arpa/inet.h>
  28 #include <netinet/in.h>
  29
  30 #include <linux/rtnetlink.h>
  31 #include <linux/sockios.h>
  32 #include <linux/ip.h>
  33 #include <linux/if_link.h>
  34 #include <linux/if_vlan.h>
  35 #include <linux/if_bridge.h>
  36 #include <linux/if_tunnel.h>
  37 #include <linux/ip6_tunnel.h>
  38 #include <linux/ethtool.h>
  39 #include <linux/fib_rules.h>
  40 #include <linux/version.h>
  41
  42 #ifndef RTN_FAILED_POLICY
  43 #define RTN_FAILED_POLICY 12
  44 #endif
  45
  46 #include <string.h>
  47 #include <fcntl.h>
  48 #include <glob.h>
  49 #include <time.h>
  50
  51 #include <netlink/msg.h>
  52 #include <netlink/attr.h>
  53 #include <netlink/socket.h>
  54 #include <libubox/uloop.h>
  55
  56 #include "netifd.h"
  57 #include "device.h"
  58 #include "system.h"
  59
  60 struct event_socket {
  61         struct uloop_fd uloop;
  62         struct nl_sock *sock;
  63         int bufsize;
  64 };
  65
  66 static int sock_ioctl = -1;
  67 static struct nl_sock *sock_rtnl = NULL;
  68
  69 static int cb_rtnl_event(struct nl_msg *msg, void *arg);
  70 static void handle_hotplug_event(struct uloop_fd *u, unsigned int events);
  71
  72 static char dev_buf[256];
  73
  74 static void
  75 handler_nl_event(struct uloop_fd *u, unsigned int events)
  76 {
  77         struct event_socket *ev = container_of(u, struct event_socket, uloop);
  78         int err;
  79         socklen_t errlen = sizeof(err);
  80
  81         if (!u->error) {
  82                 nl_recvmsgs_default(ev->sock);
  83                 return;
  84         }
  85
  86         if (getsockopt(u->fd, SOL_SOCKET, SO_ERROR, (void *)&err, &errlen))
  87                 goto abort;
  88
  89         switch(err) {
  90         case ENOBUFS:
  91                 // Increase rx buffer size on netlink socket
  92                 ev->bufsize *= 2;
  93                 if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
  94                         goto abort;
  95
  96                 // Request full dump since some info got dropped
  97                 struct rtgenmsg msg = { .rtgen_family = AF_UNSPEC };
  98                 nl_send_simple(ev->sock, RTM_GETLINK, NLM_F_DUMP, &msg, sizeof(msg));
  99                 break;
 100
 101         default:
 102                 goto abort;
 103         }
 104         u->error = false;
 105         return;
 106
 107 abort:
 108         uloop_fd_delete(&ev->uloop);
 109         return;
 110 }
 111
 112 static struct nl_sock *
 113 create_socket(int protocol, int groups)
 114 {
 115         struct nl_sock *sock;
 116
 117         sock = nl_socket_alloc();
 118         if (!sock)
 119                 return NULL;
 120
 121         if (groups)
 122                 nl_join_groups(sock, groups);
 123
 124         if (nl_connect(sock, protocol))
 125                 return NULL;
 126
 127         return sock;
 128 }
 129
 130 static bool
 131 create_raw_event_socket(struct event_socket *ev, int protocol, int groups,
 132                         uloop_fd_handler cb, int flags)
 133 {
 134         ev->sock = create_socket(protocol, groups);
 135         if (!ev->sock)
 136                 return false;
 137
 138         ev->uloop.fd = nl_socket_get_fd(ev->sock);
 139         ev->uloop.cb = cb;
 140         if (uloop_fd_add(&ev->uloop, ULOOP_READ|flags))
 141                 return false;
 142
 143         return true;
 144 }
 145
 146 static bool
 147 create_event_socket(struct event_socket *ev, int protocol,
 148                     int (*cb)(struct nl_msg *msg, void *arg))
 149 {
 150         if (!create_raw_event_socket(ev, protocol, 0, handler_nl_event, ULOOP_ERROR_CB))
 151                 return false;
 152
 153         // Install the valid custom callback handler
 154         nl_socket_modify_cb(ev->sock, NL_CB_VALID, NL_CB_CUSTOM, cb, NULL);
 155
 156         // Disable sequence number checking on event sockets
 157         nl_socket_disable_seq_check(ev->sock);
 158
 159         // Increase rx buffer size to 65K on event sockets
 160         ev->bufsize = 65535;
 161         if (nl_socket_set_buffer_size(ev->sock, ev->bufsize, 0))
 162                 return false;
 163
 164         return true;
 165 }
 166
 167 static bool
 168 system_rtn_aton(const char *src, unsigned int *dst)
 169 {
 170         char *e;
 171         unsigned int n;
 172
 173         if (!strcmp(src, "local"))
 174                 n = RTN_LOCAL;
 175         else if (!strcmp(src, "nat"))
 176                 n = RTN_NAT;
 177         else if (!strcmp(src, "broadcast"))
 178                 n = RTN_BROADCAST;
 179         else if (!strcmp(src, "anycast"))
 180                 n = RTN_ANYCAST;
 181         else if (!strcmp(src, "multicast"))
 182                 n = RTN_MULTICAST;
 183         else if (!strcmp(src, "prohibit"))
 184                 n = RTN_PROHIBIT;
 185         else if (!strcmp(src, "unreachable"))
 186                 n = RTN_UNREACHABLE;
 187         else if (!strcmp(src, "blackhole"))
 188                 n = RTN_BLACKHOLE;
 189         else if (!strcmp(src, "xresolve"))
 190                 n = RTN_XRESOLVE;
 191         else if (!strcmp(src, "unicast"))
 192                 n = RTN_UNICAST;
 193         else if (!strcmp(src, "throw"))
 194                 n = RTN_THROW;
 195         else if (!strcmp(src, "failed_policy"))
 196                 n = RTN_FAILED_POLICY;
 197         else {
 198                 n = strtoul(src, &e, 0);
 199                 if (!e || *e || e == src || n > 255)
 200                         return false;
 201         }
 202
 203         *dst = n;
 204         return true;
 205 }
 206
 207 static bool
 208 system_tos_aton(const char *src, unsigned *dst)
 209 {
 210         char *e;
 211
 212         *dst = strtoul(src, &e, 16);
 213         if (e == src || *e || *dst > 255)
 214                 return false;
 215
 216         return true;
 217 }
 218
 219 int system_init(void)
 220 {
 221         static struct event_socket rtnl_event;
 222         static struct event_socket hotplug_event;
 223
 224         sock_ioctl = socket(AF_LOCAL, SOCK_DGRAM, 0);
 225         system_fd_set_cloexec(sock_ioctl);
 226
 227         // Prepare socket for routing / address control
 228         sock_rtnl = create_socket(NETLINK_ROUTE, 0);
 229         if (!sock_rtnl)
 230                 return -1;
 231
 232         if (!create_event_socket(&rtnl_event, NETLINK_ROUTE, cb_rtnl_event))
 233                 return -1;
 234
 235         if (!create_raw_event_socket(&hotplug_event, NETLINK_KOBJECT_UEVENT, 1,
 236                                      handle_hotplug_event, 0))
 237                 return -1;
 238
 239         // Receive network link events form kernel
 240         nl_socket_add_membership(rtnl_event.sock, RTNLGRP_LINK);
 241
 242         return 0;
 243 }
 244
 245 static void system_set_sysctl(const char *path, const char *val)
 246 {
 247         int fd;
 248
 249         fd = open(path, O_WRONLY);
 250         if (fd < 0)
 251                 return;
 252
 253         if (write(fd, val, strlen(val))) {}
 254         close(fd);
 255 }
 256
 257 static void system_set_dev_sysctl(const char *path, const char *device, const char *val)
 258 {
 259         snprintf(dev_buf, sizeof(dev_buf), path, device);
 260         system_set_sysctl(dev_buf, val);
 261 }
 262
 263 static void system_set_disable_ipv6(struct device *dev, const char *val)
 264 {
 265         system_set_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6", dev->ifname, val);
 266 }
 267
 268 static int system_get_sysctl(const char *path, char *buf, const size_t buf_sz)
 269 {
 270         int fd = -1, ret = -1;
 271
 272         fd = open(path, O_RDONLY);
 273         if (fd < 0)
 274                 goto out;
 275
 276         ssize_t len = read(fd, buf, buf_sz - 1);
 277         if (len < 0)
 278                 goto out;
 279
 280         ret = buf[len] = 0;
 281
 282 out:
 283         if (fd >= 0)
 284                 close(fd);
 285
 286         return ret;
 287 }
 288
 289 static int
 290 system_get_dev_sysctl(const char *path, const char *device, char *buf, const size_t buf_sz)
 291 {
 292         snprintf(dev_buf, sizeof(dev_buf), path, device);
 293         return system_get_sysctl(dev_buf, buf, buf_sz);
 294 }
 295
 296 static int system_get_disable_ipv6(struct device *dev, char *buf, const size_t buf_sz)
 297 {
 298         return system_get_dev_sysctl("/proc/sys/net/ipv6/conf/%s/disable_ipv6",
 299                         dev->ifname, buf, buf_sz);
 300 }
 301
 302 #ifndef IFF_LOWER_UP
 303 #define IFF_LOWER_UP    0x10000
 304 #endif
 305
 306 // Evaluate netlink messages
 307 static int cb_rtnl_event(struct nl_msg *msg, void *arg)
 308 {
 309         struct nlmsghdr *nh = nlmsg_hdr(msg);
 310         struct ifinfomsg *ifi = NLMSG_DATA(nh);
 311         struct nlattr *nla[__IFLA_MAX];
 312
 313         if (nh->nlmsg_type != RTM_NEWLINK)
 314                 goto out;
 315
 316         nlmsg_parse(nh, sizeof(*ifi), nla, __IFLA_MAX - 1, NULL);
 317         if (!nla[IFLA_IFNAME])
 318                 goto out;
 319
 320         struct device *dev = device_get(nla_data(nla[IFLA_IFNAME]), false);
 321         if (!dev)
 322                 goto out;
 323
 324         device_set_ifindex(dev, ifi->ifi_index);
 325         if (!dev->type->keep_link_status)
 326                 device_set_link(dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
 327
 328 out:
 329         return 0;
 330 }
 331
 332 static void
 333 handle_hotplug_msg(char *data, int size)
 334 {
 335         const char *subsystem = NULL, *interface = NULL;
 336         char *cur, *end, *sep;
 337         struct device *dev;
 338         int skip;
 339         bool add;
 340
 341         if (!strncmp(data, "add@", 4))
 342                 add = true;
 343         else if (!strncmp(data, "remove@", 7))
 344                 add = false;
 345         else
 346                 return;
 347
 348         skip = strlen(data) + 1;
 349         end = data + size;
 350
 351         for (cur = data + skip; cur < end; cur += skip) {
 352                 skip = strlen(cur) + 1;
 353
 354                 sep = strchr(cur, '=');
 355                 if (!sep)
 356                         continue;
 357
 358                 *sep = 0;
 359                 if (!strcmp(cur, "INTERFACE"))
 360                         interface = sep + 1;
 361                 else if (!strcmp(cur, "SUBSYSTEM")) {
 362                         subsystem = sep + 1;
 363                         if (strcmp(subsystem, "net") != 0)
 364                                 return;
 365                 }
 366                 if (subsystem && interface)
 367                         goto found;
 368         }
 369         return;
 370
 371 found:
 372         dev = device_get(interface, false);
 373         if (!dev)
 374                 return;
 375
 376         if (dev->type != &simple_device_type)
 377                 return;
 378
 379         if (add && system_if_force_external(dev->ifname))
 380                 return;
 381
 382         device_set_present(dev, add);
 383 }
 384
 385 static void
 386 handle_hotplug_event(struct uloop_fd *u, unsigned int events)
 387 {
 388         struct event_socket *ev = container_of(u, struct event_socket, uloop);
 389         struct sockaddr_nl nla;
 390         unsigned char *buf = NULL;
 391         int size;
 392
 393         while ((size = nl_recv(ev->sock, &nla, &buf, NULL)) > 0) {
 394                 if (nla.nl_pid == 0)
 395                         handle_hotplug_msg((char *) buf, size);
 396
 397                 free(buf);
 398         }
 399 }
 400
 401 static int system_rtnl_call(struct nl_msg *msg)
 402 {
 403         int ret;
 404
 405         ret = nl_send_auto_complete(sock_rtnl, msg);
 406         nlmsg_free(msg);
 407
 408         if (ret < 0)
 409                 return ret;
 410
 411         return nl_wait_for_ack(sock_rtnl);
 412 }
 413
 414 int system_bridge_delbr(struct device *bridge)
 415 {
 416         return ioctl(sock_ioctl, SIOCBRDELBR, bridge->ifname);
 417 }
 418
 419 static int system_bridge_if(const char *bridge, struct device *dev, int cmd, void *data)
 420 {
 421         struct ifreq ifr;
 422
 423         memset(&ifr, 0, sizeof(ifr));
 424         if (dev)
 425                 ifr.ifr_ifindex = dev->ifindex;
 426         else
 427                 ifr.ifr_data = data;
 428         strncpy(ifr.ifr_name, bridge, sizeof(ifr.ifr_name));
 429         return ioctl(sock_ioctl, cmd, &ifr);
 430 }
 431
 432 static bool system_is_bridge(const char *name, char *buf, int buflen)
 433 {
 434         struct stat st;
 435
 436         snprintf(buf, buflen, "/sys/devices/virtual/net/%s/bridge", name);
 437         if (stat(buf, &st) < 0)
 438                 return false;
 439
 440         return true;
 441 }
 442
 443 static char *system_get_bridge(const char *name, char *buf, int buflen)
 444 {
 445         char *path;
 446         ssize_t len = -1;
 447         glob_t gl;
 448
 449         snprintf(buf, buflen, "/sys/devices/virtual/net/*/brif/%s/bridge", name);
 450         if (glob(buf, GLOB_NOSORT, NULL, &gl) < 0)
 451                 return NULL;
 452
 453         if (gl.gl_pathc > 0)
 454                 len = readlink(gl.gl_pathv[0], buf, buflen);
 455
 456         globfree(&gl);
 457
 458         if (len < 0)
 459                 return NULL;
 460
 461         buf[len] = 0;
 462         path = strrchr(buf, '/');
 463         if (!path)
 464                 return NULL;
 465
 466         return path + 1;
 467 }
 468
 469 static void system_bridge_set_wireless(const char *bridge, const char *dev)
 470 {
 471         snprintf(dev_buf, sizeof(dev_buf),
 472                  "/sys/devices/virtual/net/%s/brif/%s/multicast_to_unicast",
 473                  bridge, dev);
 474         system_set_sysctl(dev_buf, "1");
 475 }
 476
 477 int system_bridge_addif(struct device *bridge, struct device *dev)
 478 {
 479         char *oldbr;
 480         int ret = 0;
 481
 482         oldbr = system_get_bridge(dev->ifname, dev_buf, sizeof(dev_buf));
 483         if (!oldbr || strcmp(oldbr, bridge->ifname) != 0)
 484                 ret = system_bridge_if(bridge->ifname, dev, SIOCBRADDIF, NULL);
 485
 486         if (dev->wireless)
 487                 system_bridge_set_wireless(bridge->ifname, dev->ifname);
 488
 489         return ret;
 490 }
 491
 492 int system_bridge_delif(struct device *bridge, struct device *dev)
 493 {
 494         return system_bridge_if(bridge->ifname, dev, SIOCBRDELIF, NULL);
 495 }
 496
 497 static int system_if_resolve(struct device *dev)
 498 {
 499         struct ifreq ifr;
 500         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 501         if (!ioctl(sock_ioctl, SIOCGIFINDEX, &ifr))
 502                 return ifr.ifr_ifindex;
 503         else
 504                 return 0;
 505 }
 506
 507 static int system_if_flags(const char *ifname, unsigned add, unsigned rem)
 508 {
 509         struct ifreq ifr;
 510
 511         memset(&ifr, 0, sizeof(ifr));
 512         strncpy(ifr.ifr_name, ifname, sizeof(ifr.ifr_name));
 513         ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr);
 514         ifr.ifr_flags |= add;
 515         ifr.ifr_flags &= ~rem;
 516         return ioctl(sock_ioctl, SIOCSIFFLAGS, &ifr);
 517 }
 518
 519 struct clear_data {
 520         struct nl_msg *msg;
 521         struct device *dev;
 522         int type;
 523         int size;
 524         int af;
 525 };
 526
 527
 528 static bool check_ifaddr(struct nlmsghdr *hdr, int ifindex)
 529 {
 530         struct ifaddrmsg *ifa = NLMSG_DATA(hdr);
 531
 532         return ifa->ifa_index == ifindex;
 533 }
 534
 535 static bool check_route(struct nlmsghdr *hdr, int ifindex)
 536 {
 537         struct rtmsg *r = NLMSG_DATA(hdr);
 538         struct nlattr *tb[__RTA_MAX];
 539
 540         if (r->rtm_protocol == RTPROT_KERNEL &&
 541             r->rtm_family == AF_INET6)
 542                 return false;
 543
 544         nlmsg_parse(hdr, sizeof(struct rtmsg), tb, __RTA_MAX - 1, NULL);
 545         if (!tb[RTA_OIF])
 546                 return false;
 547
 548         return *(int *)RTA_DATA(tb[RTA_OIF]) == ifindex;
 549 }
 550
 551 static bool check_rule(struct nlmsghdr *hdr, int ifindex)
 552 {
 553         return true;
 554 }
 555
 556 static int cb_clear_event(struct nl_msg *msg, void *arg)
 557 {
 558         struct clear_data *clr = arg;
 559         struct nlmsghdr *hdr = nlmsg_hdr(msg);
 560         bool (*cb)(struct nlmsghdr *, int ifindex);
 561         int type;
 562
 563         switch(clr->type) {
 564         case RTM_GETADDR:
 565                 type = RTM_DELADDR;
 566                 if (hdr->nlmsg_type != RTM_NEWADDR)
 567                         return NL_SKIP;
 568
 569                 cb = check_ifaddr;
 570                 break;
 571         case RTM_GETROUTE:
 572                 type = RTM_DELROUTE;
 573                 if (hdr->nlmsg_type != RTM_NEWROUTE)
 574                         return NL_SKIP;
 575
 576                 cb = check_route;
 577                 break;
 578         case RTM_GETRULE:
 579                 type = RTM_DELRULE;
 580                 if (hdr->nlmsg_type != RTM_NEWRULE)
 581                         return NL_SKIP;
 582
 583                 cb = check_rule;
 584                 break;
 585         default:
 586                 return NL_SKIP;
 587         }
 588
 589         if (!cb(hdr, clr->dev ? clr->dev->ifindex : 0))
 590                 return NL_SKIP;
 591
 592         if (type == RTM_DELRULE)
 593                 D(SYSTEM, "Remove a rule\n");
 594         else
 595                 D(SYSTEM, "Remove %s from device %s\n",
 596                   type == RTM_DELADDR ? "an address" : "a route",
 597                   clr->dev->ifname);
 598         memcpy(nlmsg_hdr(clr->msg), hdr, hdr->nlmsg_len);
 599         hdr = nlmsg_hdr(clr->msg);
 600         hdr->nlmsg_type = type;
 601         hdr->nlmsg_flags = NLM_F_REQUEST;
 602
 603         nl_socket_disable_auto_ack(sock_rtnl);
 604         nl_send_auto_complete(sock_rtnl, clr->msg);
 605         nl_socket_enable_auto_ack(sock_rtnl);
 606
 607         return NL_SKIP;
 608 }
 609
 610 static int
 611 cb_finish_event(struct nl_msg *msg, void *arg)
 612 {
 613         int *pending = arg;
 614         *pending = 0;
 615         return NL_STOP;
 616 }
 617
 618 static int
 619 error_handler(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
 620 {
 621         int *pending = arg;
 622         *pending = err->error;
 623         return NL_STOP;
 624 }
 625
 626 static void
 627 system_if_clear_entries(struct device *dev, int type, int af)
 628 {
 629         struct clear_data clr;
 630         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
 631         struct rtmsg rtm = {
 632                 .rtm_family = af,
 633                 .rtm_flags = RTM_F_CLONED,
 634         };
 635         int flags = NLM_F_DUMP;
 636         int pending = 1;
 637
 638         clr.af = af;
 639         clr.dev = dev;
 640         clr.type = type;
 641         switch (type) {
 642         case RTM_GETADDR:
 643         case RTM_GETRULE:
 644                 clr.size = sizeof(struct rtgenmsg);
 645                 break;
 646         case RTM_GETROUTE:
 647                 clr.size = sizeof(struct rtmsg);
 648                 break;
 649         default:
 650                 return;
 651         }
 652
 653         if (!cb)
 654                 return;
 655
 656         clr.msg = nlmsg_alloc_simple(type, flags);
 657         if (!clr.msg)
 658                 goto out;
 659
 660         nlmsg_append(clr.msg, &rtm, clr.size, 0);
 661         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_clear_event, &clr);
 662         nl_cb_set(cb, NL_CB_FINISH, NL_CB_CUSTOM, cb_finish_event, &pending);
 663         nl_cb_err(cb, NL_CB_CUSTOM, error_handler, &pending);
 664
 665         nl_send_auto_complete(sock_rtnl, clr.msg);
 666         while (pending > 0)
 667                 nl_recvmsgs(sock_rtnl, cb);
 668
 669         nlmsg_free(clr.msg);
 670 out:
 671         nl_cb_put(cb);
 672 }
 673
 674 /*
 675  * Clear bridge (membership) state and bring down device
 676  */
 677 void system_if_clear_state(struct device *dev)
 678 {
 679         static char buf[256];
 680         char *bridge;
 681
 682         device_set_ifindex(dev, system_if_resolve(dev));
 683         if (dev->external || !dev->ifindex)
 684                 return;
 685
 686         system_if_flags(dev->ifname, 0, IFF_UP);
 687
 688         if (system_is_bridge(dev->ifname, buf, sizeof(buf))) {
 689                 D(SYSTEM, "Delete existing bridge named '%s'\n", dev->ifname);
 690                 system_bridge_delbr(dev);
 691                 return;
 692         }
 693
 694         bridge = system_get_bridge(dev->ifname, buf, sizeof(buf));
 695         if (bridge) {
 696                 D(SYSTEM, "Remove device '%s' from bridge '%s'\n", dev->ifname, bridge);
 697                 system_bridge_if(bridge, dev, SIOCBRDELIF, NULL);
 698         }
 699
 700         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET);
 701         system_if_clear_entries(dev, RTM_GETADDR, AF_INET);
 702         system_if_clear_entries(dev, RTM_GETROUTE, AF_INET6);
 703         system_if_clear_entries(dev, RTM_GETADDR, AF_INET6);
 704         system_set_disable_ipv6(dev, "0");
 705 }
 706
 707 static inline unsigned long
 708 sec_to_jiffies(int val)
 709 {
 710         return (unsigned long) val * 100;
 711 }
 712
 713 int system_bridge_addbr(struct device *bridge, struct bridge_config *cfg)
 714 {
 715         unsigned long args[4] = {};
 716
 717         if (ioctl(sock_ioctl, SIOCBRADDBR, bridge->ifname) < 0)
 718                 return -1;
 719
 720         args[0] = BRCTL_SET_BRIDGE_STP_STATE;
 721         args[1] = !!cfg->stp;
 722         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 723
 724         args[0] = BRCTL_SET_BRIDGE_FORWARD_DELAY;
 725         args[1] = sec_to_jiffies(cfg->forward_delay);
 726         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 727
 728         system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_snooping",
 729                 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 730
 731         system_set_dev_sysctl("/sys/devices/virtual/net/%s/bridge/multicast_querier",
 732                 bridge->ifname, cfg->igmp_snoop ? "1" : "0");
 733
 734         args[0] = BRCTL_SET_BRIDGE_PRIORITY;
 735         args[1] = cfg->priority;
 736         system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 737
 738         if (cfg->flags & BRIDGE_OPT_AGEING_TIME) {
 739                 args[0] = BRCTL_SET_AGEING_TIME;
 740                 args[1] = sec_to_jiffies(cfg->ageing_time);
 741                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 742         }
 743
 744         if (cfg->flags & BRIDGE_OPT_HELLO_TIME) {
 745                 args[0] = BRCTL_SET_BRIDGE_HELLO_TIME;
 746                 args[1] = sec_to_jiffies(cfg->hello_time);
 747                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 748         }
 749
 750         if (cfg->flags & BRIDGE_OPT_MAX_AGE) {
 751                 args[0] = BRCTL_SET_BRIDGE_MAX_AGE;
 752                 args[1] = sec_to_jiffies(cfg->max_age);
 753                 system_bridge_if(bridge->ifname, NULL, SIOCDEVPRIVATE, &args);
 754         }
 755
 756         return 0;
 757 }
 758
 759 int system_macvlan_add(struct device *macvlan, struct device *dev, struct macvlan_config *cfg)
 760 {
 761         struct nl_msg *msg;
 762         struct nlattr *linkinfo, *data;
 763         struct ifinfomsg iim = { .ifi_family = AF_UNSPEC, };
 764         int ifindex = system_if_resolve(dev);
 765         int i, rv;
 766         static const struct {
 767                 const char *name;
 768                 enum macvlan_mode val;
 769         } modes[] = {
 770                 { "private", MACVLAN_MODE_PRIVATE },
 771                 { "vepa", MACVLAN_MODE_VEPA },
 772                 { "bridge", MACVLAN_MODE_BRIDGE },
 773                 { "passthru", MACVLAN_MODE_PASSTHRU },
 774         };
 775
 776         if (ifindex == 0)
 777                 return -ENOENT;
 778
 779         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 780
 781         if (!msg)
 782                 return -1;
 783
 784         nlmsg_append(msg, &iim, sizeof(iim), 0);
 785
 786         if (cfg->flags & MACVLAN_OPT_MACADDR)
 787                 nla_put(msg, IFLA_ADDRESS, sizeof(cfg->macaddr), cfg->macaddr);
 788         nla_put_string(msg, IFLA_IFNAME, macvlan->ifname);
 789         nla_put_u32(msg, IFLA_LINK, ifindex);
 790
 791         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 792                 goto nla_put_failure;
 793
 794         nla_put_string(msg, IFLA_INFO_KIND, "macvlan");
 795
 796         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 797                 goto nla_put_failure;
 798
 799         if (cfg->mode) {
 800                 for (i = 0; i < ARRAY_SIZE(modes); i++) {
 801                         if (strcmp(cfg->mode, modes[i].name) != 0)
 802                                 continue;
 803
 804                         nla_put_u32(msg, IFLA_MACVLAN_MODE, modes[i].val);
 805                         break;
 806                 }
 807         }
 808
 809         nla_nest_end(msg, data);
 810         nla_nest_end(msg, linkinfo);
 811
 812         rv = system_rtnl_call(msg);
 813         if (rv)
 814                 D(SYSTEM, "Error adding macvlan '%s' over '%s': %d\n", macvlan->ifname, dev->ifname, rv);
 815
 816         return rv;
 817
 818 nla_put_failure:
 819         nlmsg_free(msg);
 820         return -ENOMEM;
 821 }
 822
 823 static int system_link_del(const char *ifname)
 824 {
 825         struct nl_msg *msg;
 826         struct ifinfomsg iim = {
 827                 .ifi_family = AF_UNSPEC,
 828                 .ifi_index = 0,
 829         };
 830
 831         msg = nlmsg_alloc_simple(RTM_DELLINK, NLM_F_REQUEST);
 832
 833         if (!msg)
 834                 return -1;
 835
 836         nlmsg_append(msg, &iim, sizeof(iim), 0);
 837         nla_put_string(msg, IFLA_IFNAME, ifname);
 838         return system_rtnl_call(msg);
 839 }
 840
 841 int system_macvlan_del(struct device *macvlan)
 842 {
 843         return system_link_del(macvlan->ifname);
 844 }
 845
 846 static int system_vlan(struct device *dev, int id)
 847 {
 848         struct vlan_ioctl_args ifr = {
 849                 .cmd = SET_VLAN_NAME_TYPE_CMD,
 850                 .u.name_type = VLAN_NAME_TYPE_RAW_PLUS_VID_NO_PAD,
 851         };
 852
 853         ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 854
 855         if (id < 0) {
 856                 ifr.cmd = DEL_VLAN_CMD;
 857                 ifr.u.VID = 0;
 858         } else {
 859                 ifr.cmd = ADD_VLAN_CMD;
 860                 ifr.u.VID = id;
 861         }
 862         strncpy(ifr.device1, dev->ifname, sizeof(ifr.device1));
 863         return ioctl(sock_ioctl, SIOCSIFVLAN, &ifr);
 864 }
 865
 866 int system_vlan_add(struct device *dev, int id)
 867 {
 868         return system_vlan(dev, id);
 869 }
 870
 871 int system_vlan_del(struct device *dev)
 872 {
 873         return system_vlan(dev, -1);
 874 }
 875
 876 int system_vlandev_add(struct device *vlandev, struct device *dev, struct vlandev_config *cfg)
 877 {
 878         struct nl_msg *msg;
 879         struct nlattr *linkinfo, *data;
 880         struct ifinfomsg iim = { .ifi_family = AF_UNSPEC };
 881         int ifindex = system_if_resolve(dev);
 882         int rv;
 883
 884         if (ifindex == 0)
 885                 return -ENOENT;
 886
 887         msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
 888
 889         if (!msg)
 890                 return -1;
 891
 892         nlmsg_append(msg, &iim, sizeof(iim), 0);
 893         nla_put_string(msg, IFLA_IFNAME, vlandev->ifname);
 894         nla_put_u32(msg, IFLA_LINK, ifindex);
 895
 896         if (!(linkinfo = nla_nest_start(msg, IFLA_LINKINFO)))
 897                 goto nla_put_failure;
 898
 899         nla_put_string(msg, IFLA_INFO_KIND, "vlan");
 900
 901         if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
 902                 goto nla_put_failure;
 903
 904         nla_put_u16(msg, IFLA_VLAN_ID, cfg->vid);
 905
 906 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,10,0)
 907         nla_put_u16(msg, IFLA_VLAN_PROTOCOL, htons(cfg->proto));
 908 #else
 909         if(cfg->proto == VLAN_PROTO_8021AD)
 910                 netifd_log_message(L_WARNING, "%s Your kernel is older than linux 3.10.0, 802.1ad is not supported defaulting to 802.1q", vlandev->type->name);
 911 #endif
 912
 913         nla_nest_end(msg, data);
 914         nla_nest_end(msg, linkinfo);
 915
 916         rv = system_rtnl_call(msg);
 917         if (rv)
 918                 D(SYSTEM, "Error adding vlandev '%s' over '%s': %d\n", vlandev->ifname, dev->ifname, rv);
 919
 920         return rv;
 921
 922 nla_put_failure:
 923         nlmsg_free(msg);
 924         return -ENOMEM;
 925 }
 926
 927 int system_vlandev_del(struct device *vlandev)
 928 {
 929         return system_link_del(vlandev->ifname);
 930 }
 931
 932 static void
 933 system_if_get_settings(struct device *dev, struct device_settings *s)
 934 {
 935         struct ifreq ifr;
 936         char buf[10];
 937
 938         memset(&ifr, 0, sizeof(ifr));
 939         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 940
 941         if (ioctl(sock_ioctl, SIOCGIFMTU, &ifr) == 0) {
 942                 s->mtu = ifr.ifr_mtu;
 943                 s->flags |= DEV_OPT_MTU;
 944         }
 945
 946         if (ioctl(sock_ioctl, SIOCGIFTXQLEN, &ifr) == 0) {
 947                 s->txqueuelen = ifr.ifr_qlen;
 948                 s->flags |= DEV_OPT_TXQUEUELEN;
 949         }
 950
 951         if (ioctl(sock_ioctl, SIOCGIFHWADDR, &ifr) == 0) {
 952                 memcpy(s->macaddr, &ifr.ifr_hwaddr.sa_data, sizeof(s->macaddr));
 953                 s->flags |= DEV_OPT_MACADDR;
 954         }
 955
 956         if (!system_get_disable_ipv6(dev, buf, sizeof(buf))) {
 957                 s->ipv6 = !strtoul(buf, NULL, 0);
 958                 s->flags |= DEV_OPT_IPV6;
 959         }
 960
 961         if (ioctl(sock_ioctl, SIOCGIFFLAGS, &ifr) == 0) {
 962                 s->promisc = ifr.ifr_flags & IFF_PROMISC;
 963                 s->flags |= DEV_OPT_PROMISC;
 964         }
 965 }
 966
 967 void
 968 system_if_apply_settings(struct device *dev, struct device_settings *s, unsigned int apply_mask)
 969 {
 970         struct ifreq ifr;
 971
 972         if (!apply_mask)
 973                 return;
 974
 975         memset(&ifr, 0, sizeof(ifr));
 976         strncpy(ifr.ifr_name, dev->ifname, sizeof(ifr.ifr_name));
 977         if (s->flags & DEV_OPT_MTU & apply_mask) {
 978                 ifr.ifr_mtu = s->mtu;
 979                 if (ioctl(sock_ioctl, SIOCSIFMTU, &ifr) < 0)
 980                         s->flags &= ~DEV_OPT_MTU;
 981         }
 982         if (s->flags & DEV_OPT_TXQUEUELEN & apply_mask) {
 983                 ifr.ifr_qlen = s->txqueuelen;
 984                 if (ioctl(sock_ioctl, SIOCSIFTXQLEN, &ifr) < 0)
 985                         s->flags &= ~DEV_OPT_TXQUEUELEN;
 986         }
 987         if ((s->flags & DEV_OPT_MACADDR & apply_mask) && !dev->external) {
 988                 ifr.ifr_hwaddr.sa_family = ARPHRD_ETHER;
 989                 memcpy(&ifr.ifr_hwaddr.sa_data, s->macaddr, sizeof(s->macaddr));
 990                 if (ioctl(sock_ioctl, SIOCSIFHWADDR, &ifr) < 0)
 991                         s->flags &= ~DEV_OPT_MACADDR;
 992         }
 993         if (s->flags & DEV_OPT_IPV6 & apply_mask)
 994                 system_set_disable_ipv6(dev, s->ipv6 ? "0" : "1");
 995         if (s->flags & DEV_OPT_PROMISC & apply_mask) {
 996                 if (system_if_flags(dev->ifname, s->promisc ? IFF_PROMISC : 0,
 997                                     !s->promisc ? IFF_PROMISC : 0) < 0)
 998                         s->flags &= ~DEV_OPT_PROMISC;
 999         }
1000 }
1001
1002 int system_if_up(struct device *dev)
1003 {
1004         system_if_get_settings(dev, &dev->orig_settings);
1005         system_if_apply_settings(dev, &dev->settings, dev->settings.flags);
1006         device_set_ifindex(dev, system_if_resolve(dev));
1007         return system_if_flags(dev->ifname, IFF_UP, 0);
1008 }
1009
1010 int system_if_down(struct device *dev)
1011 {
1012         int ret = system_if_flags(dev->ifname, 0, IFF_UP);
1013         dev->orig_settings.flags &= dev->settings.flags;
1014         system_if_apply_settings(dev, &dev->orig_settings, dev->orig_settings.flags);
1015         return ret;
1016 }
1017
1018 struct if_check_data {
1019         struct device *dev;
1020         int pending;
1021         int ret;
1022 };
1023
1024 static int cb_if_check_valid(struct nl_msg *msg, void *arg)
1025 {
1026         struct nlmsghdr *nh = nlmsg_hdr(msg);
1027         struct ifinfomsg *ifi = NLMSG_DATA(nh);
1028         struct if_check_data *chk = (struct if_check_data *)arg;
1029
1030         if (nh->nlmsg_type != RTM_NEWLINK)
1031                 return NL_SKIP;
1032
1033         device_set_present(chk->dev, ifi->ifi_index > 0 ? true : false);
1034         device_set_link(chk->dev, ifi->ifi_flags & IFF_LOWER_UP ? true : false);
1035
1036         return NL_OK;
1037 }
1038
1039 static int cb_if_check_ack(struct nl_msg *msg, void *arg)
1040 {
1041         struct if_check_data *chk = (struct if_check_data *)arg;
1042         chk->pending = 0;
1043         return NL_STOP;
1044 }
1045
1046 static int cb_if_check_error(struct sockaddr_nl *nla, struct nlmsgerr *err, void *arg)
1047 {
1048         struct if_check_data *chk = (struct if_check_data *)arg;
1049
1050         device_set_present(chk->dev, false);
1051         device_set_link(chk->dev, false);
1052         chk->pending = err->error;
1053
1054         return NL_STOP;
1055 }
1056
1057 int system_if_check(struct device *dev)
1058 {
1059         struct nl_cb *cb = nl_cb_alloc(NL_CB_DEFAULT);
1060         struct nl_msg *msg;
1061         struct ifinfomsg ifi = {
1062                 .ifi_family = AF_UNSPEC,
1063                 .ifi_index = 0,
1064         };
1065         struct if_check_data chk = {
1066                 .dev = dev,
1067                 .pending = 1,
1068         };
1069         int ret = 1;
1070
1071         msg = nlmsg_alloc_simple(RTM_GETLINK, 0);
1072         if (!msg || nlmsg_append(msg, &ifi, sizeof(ifi), 0) ||
1073             nla_put_string(msg, IFLA_IFNAME, dev->ifname))
1074                 goto out;
1075
1076         nl_cb_set(cb, NL_CB_VALID, NL_CB_CUSTOM, cb_if_check_valid, &chk);
1077         nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, cb_if_check_ack, &chk);
1078         nl_cb_err(cb, NL_CB_CUSTOM, cb_if_check_error, &chk);
1079
1080         nl_send_auto_complete(sock_rtnl, msg);
1081         while (chk.pending > 0)
1082                 nl_recvmsgs(sock_rtnl, cb);
1083
1084         nlmsg_free(msg);
1085         ret = chk.pending;
1086
1087 out:
1088         nl_cb_put(cb);
1089         return ret;
1090 }
1091
1092 struct device *
1093 system_if_get_parent(struct device *dev)
1094 {
1095         char buf[64], *devname;
1096         int ifindex, iflink, len;
1097         FILE *f;
1098
1099         snprintf(buf, sizeof(buf), "/sys/class/net/%s/iflink", dev->ifname);
1100         f = fopen(buf, "r");
1101         if (!f)
1102                 return NULL;
1103
1104         len = fread(buf, 1, sizeof(buf) - 1, f);
1105         fclose(f);
1106
1107         if (len <= 0)
1108                 return NULL;
1109
1110         buf[len] = 0;
1111         iflink = strtoul(buf, NULL, 0);
1112         ifindex = system_if_resolve(dev);
1113         if (!iflink || iflink == ifindex)
1114                 return NULL;
1115
1116         devname = if_indextoname(iflink, buf);
1117         if (!devname)
1118                 return NULL;
1119
1120         return device_get(devname, true);
1121 }
1122
1123 static bool
1124 read_string_file(int dir_fd, const char *file, char *buf, int len)
1125 {
1126         bool ret = false;
1127         char *c;
1128         int fd;
1129
1130         fd = openat(dir_fd, file, O_RDONLY);
1131         if (fd < 0)
1132                 return false;
1133
1134 retry:
1135         len = read(fd, buf, len - 1);
1136         if (len < 0) {
1137                 if (errno == EINTR)
1138                         goto retry;
1139         } else if (len > 0) {
1140                         buf[len] = 0;
1141
1142                         c = strchr(buf, '\n');
1143                         if (c)
1144                                 *c = 0;
1145
1146                         ret = true;
1147         }
1148
1149         close(fd);
1150
1151         return ret;
1152 }
1153
1154 static bool
1155 read_uint64_file(int dir_fd, const char *file, uint64_t *val)
1156 {
1157         char buf[64];
1158         bool ret = false;
1159
1160         ret = read_string_file(dir_fd, file, buf, sizeof(buf));
1161         if (ret)
1162                 *val = strtoull(buf, NULL, 0);
1163
1164         return ret;
1165 }
1166
1167 /* Assume advertised flags == supported flags */
1168 static const struct {
1169         uint32_t mask;
1170         const char *name;
1171 } ethtool_link_modes[] = {
1172         { ADVERTISED_10baseT_Half, "10H" },
1173         { ADVERTISED_10baseT_Full, "10F" },
1174         { ADVERTISED_100baseT_Half, "100H" },
1175         { ADVERTISED_100baseT_Full, "100F" },
1176         { ADVERTISED_1000baseT_Half, "1000H" },
1177         { ADVERTISED_1000baseT_Full, "1000F" },
1178 };
1179
1180 static void system_add_link_modes(struct blob_buf *b, __u32 mask)
1181 {
1182         int i;
1183         for (i = 0; i < ARRAY_SIZE(ethtool_link_modes); i++) {
1184                 if (mask & ethtool_link_modes[i].mask)
1185                         blobmsg_add_string(b, NULL, ethtool_link_modes[i].name);
1186         }
1187 }
1188
1189 bool
1190 system_if_force_external(const char *ifname)
1191 {
1192         char buf[64];
1193         struct stat s;
1194
1195         snprintf(buf, sizeof(buf), "/sys/class/net/%s/phy80211", ifname);
1196         return stat(buf, &s) == 0;
1197 }
1198
1199 int
1200 system_if_dump_info(struct device *dev, struct blob_buf *b)
1201 {
1202         struct ethtool_cmd ecmd;
1203         struct ifreq ifr;
1204         char buf[64], *s;
1205         void *c;
1206         int dir_fd;
1207
1208         snprintf(buf, sizeof(buf), "/sys/class/net/%s", dev->ifname);
1209         dir_fd = open(buf, O_DIRECTORY);
1210
1211         memset(&ecmd, 0, sizeof(ecmd));
1212         memset(&ifr, 0, sizeof(ifr));
1213         strcpy(ifr.ifr_name, dev->ifname);
1214         ifr.ifr_data = (caddr_t) &ecmd;
1215         ecmd.cmd = ETHTOOL_GSET;
1216
1217         if (ioctl(sock_ioctl, SIOCETHTOOL, &ifr) == 0) {
1218                 c = blobmsg_open_array(b, "link-advertising");
1219                 system_add_link_modes(b, ecmd.advertising);
1220                 blobmsg_close_array(b, c);
1221
1222                 c = blobmsg_open_array(b, "link-supported");
1223                 system_add_link_modes(b, ecmd.supported);
1224                 blobmsg_close_array(b, c);
1225
1226                 s = blobmsg_alloc_string_buffer(b, "speed", 8);
1227                 snprintf(s, 8, "%d%c", ethtool_cmd_speed(&ecmd),
1228                         ecmd.duplex == DUPLEX_HALF ? 'H' : 'F');
1229                 blobmsg_add_string_buffer(b);
1230         }
1231
1232         close(dir_fd);
1233         return 0;
1234 }
1235
1236 int
1237 system_if_dump_stats(struct device *dev, struct blob_buf *b)
1238 {
1239         const char *const counters[] = {
1240                 "collisions",     "rx_frame_errors",   "tx_compressed",
1241                 "multicast",      "rx_length_errors",  "tx_dropped",
1242                 "rx_bytes",       "rx_missed_errors",  "tx_errors",
1243                 "rx_compressed",  "rx_over_errors",    "tx_fifo_errors",
1244                 "rx_crc_errors",  "rx_packets",        "tx_heartbeat_errors",
1245                 "rx_dropped",     "tx_aborted_errors", "tx_packets",
1246                 "rx_errors",      "tx_bytes",          "tx_window_errors",
1247                 "rx_fifo_errors", "tx_carrier_errors",
1248         };
1249         char buf[64];
1250         int stats_dir;
1251         int i;
1252         uint64_t val = 0;
1253
1254         snprintf(buf, sizeof(buf), "/sys/class/net/%s/statistics", dev->ifname);
1255         stats_dir = open(buf, O_DIRECTORY);
1256         if (stats_dir < 0)
1257                 return -1;
1258
1259         for (i = 0; i < ARRAY_SIZE(counters); i++)
1260                 if (read_uint64_file(stats_dir, counters[i], &val))
1261                         blobmsg_add_u64(b, counters[i], val);
1262
1263         close(stats_dir);
1264         return 0;
1265 }
1266
1267 static int system_addr(struct device *dev, struct device_addr *addr, int cmd)
1268 {
1269         bool v4 = ((addr->flags & DEVADDR_FAMILY) == DEVADDR_INET4);
1270         int alen = v4 ? 4 : 16;
1271         unsigned int flags = 0;
1272         struct ifaddrmsg ifa = {
1273                 .ifa_family = (alen == 4) ? AF_INET : AF_INET6,
1274                 .ifa_prefixlen = addr->mask,
1275                 .ifa_index = dev->ifindex,
1276         };
1277
1278         struct nl_msg *msg;
1279         if (cmd == RTM_NEWADDR)
1280                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1281
1282         msg = nlmsg_alloc_simple(cmd, flags);
1283         if (!msg)
1284                 return -1;
1285
1286         nlmsg_append(msg, &ifa, sizeof(ifa), 0);
1287         nla_put(msg, IFA_LOCAL, alen, &addr->addr);
1288         if (v4) {
1289                 if (addr->broadcast)
1290                         nla_put_u32(msg, IFA_BROADCAST, addr->broadcast);
1291                 if (addr->point_to_point)
1292                         nla_put_u32(msg, IFA_ADDRESS, addr->point_to_point);
1293         } else {
1294                 time_t now = system_get_rtime();
1295                 struct ifa_cacheinfo cinfo = {0xffffffffU, 0xffffffffU, 0, 0};
1296
1297                 if (addr->preferred_until) {
1298                         int64_t preferred = addr->preferred_until - now;
1299                         if (preferred < 0)
1300                                 preferred = 0;
1301                         else if (preferred > UINT32_MAX)
1302                                 preferred = UINT32_MAX;
1303
1304                         cinfo.ifa_prefered = preferred;
1305                 }
1306
1307                 if (addr->valid_until) {
1308                         int64_t valid = addr->valid_until - now;
1309                         if (valid <= 0)
1310                                 return -1;
1311                         else if (valid > UINT32_MAX)
1312                                 valid = UINT32_MAX;
1313
1314                         cinfo.ifa_valid = valid;
1315                 }
1316
1317                 nla_put(msg, IFA_CACHEINFO, sizeof(cinfo), &cinfo);
1318         }
1319
1320         return system_rtnl_call(msg);
1321 }
1322
1323 int system_add_address(struct device *dev, struct device_addr *addr)
1324 {
1325         return system_addr(dev, addr, RTM_NEWADDR);
1326 }
1327
1328 int system_del_address(struct device *dev, struct device_addr *addr)
1329 {
1330         return system_addr(dev, addr, RTM_DELADDR);
1331 }
1332
1333 static int system_rt(struct device *dev, struct device_route *route, int cmd)
1334 {
1335         int alen = ((route->flags & DEVADDR_FAMILY) == DEVADDR_INET4) ? 4 : 16;
1336         bool have_gw;
1337         unsigned int flags = 0;
1338
1339         if (alen == 4)
1340                 have_gw = !!route->nexthop.in.s_addr;
1341         else
1342                 have_gw = route->nexthop.in6.s6_addr32[0] ||
1343                         route->nexthop.in6.s6_addr32[1] ||
1344                         route->nexthop.in6.s6_addr32[2] ||
1345                         route->nexthop.in6.s6_addr32[3];
1346
1347         unsigned int table = (route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))
1348                         ? route->table : RT_TABLE_MAIN;
1349
1350         struct rtmsg rtm = {
1351                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1352                 .rtm_dst_len = route->mask,
1353                 .rtm_src_len = route->sourcemask,
1354                 .rtm_table = (table < 256) ? table : RT_TABLE_UNSPEC,
1355                 .rtm_protocol = (route->flags & DEVADDR_KERNEL) ? RTPROT_KERNEL : RTPROT_STATIC,
1356                 .rtm_scope = RT_SCOPE_NOWHERE,
1357                 .rtm_type = (cmd == RTM_DELROUTE) ? 0: RTN_UNICAST,
1358                 .rtm_flags = (route->flags & DEVROUTE_ONLINK) ? RTNH_F_ONLINK : 0,
1359         };
1360         struct nl_msg *msg;
1361
1362         if (cmd == RTM_NEWROUTE) {
1363                 flags |= NLM_F_CREATE | NLM_F_REPLACE;
1364
1365                 if (!dev) { // Add null-route
1366                         rtm.rtm_scope = RT_SCOPE_UNIVERSE;
1367                         rtm.rtm_type = RTN_UNREACHABLE;
1368                 }
1369                 else
1370                         rtm.rtm_scope = (have_gw) ? RT_SCOPE_UNIVERSE : RT_SCOPE_LINK;
1371         }
1372
1373         if (route->flags & DEVROUTE_TYPE) {
1374                 rtm.rtm_type = route->type;
1375                 if (!(route->flags & (DEVROUTE_TABLE | DEVROUTE_SRCTABLE))) {
1376                         if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_BROADCAST ||
1377                             rtm.rtm_type == RTN_NAT || rtm.rtm_type == RTN_ANYCAST)
1378                                 rtm.rtm_table = RT_TABLE_LOCAL;
1379                 }
1380
1381                 if (rtm.rtm_type == RTN_LOCAL || rtm.rtm_type == RTN_NAT)
1382                         rtm.rtm_scope = RT_SCOPE_HOST;
1383                 else if (rtm.rtm_type == RTN_BROADCAST || rtm.rtm_type == RTN_MULTICAST ||
1384                         rtm.rtm_type == RTN_ANYCAST)
1385                         rtm.rtm_scope = RT_SCOPE_LINK;
1386         }
1387
1388         msg = nlmsg_alloc_simple(cmd, flags);
1389         if (!msg)
1390                 return -1;
1391
1392         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1393
1394         if (route->mask)
1395                 nla_put(msg, RTA_DST, alen, &route->addr);
1396
1397         if (route->sourcemask)
1398                 nla_put(msg, RTA_SRC, alen, &route->source);
1399
1400         if (route->metric > 0)
1401                 nla_put_u32(msg, RTA_PRIORITY, route->metric);
1402
1403         if (have_gw)
1404                 nla_put(msg, RTA_GATEWAY, alen, &route->nexthop);
1405
1406         if (dev)
1407                 nla_put_u32(msg, RTA_OIF, dev->ifindex);
1408
1409         if (table >= 256)
1410                 nla_put_u32(msg, RTA_TABLE, table);
1411
1412         if (route->flags & DEVROUTE_MTU) {
1413                 struct nlattr *metrics;
1414
1415                 if (!(metrics = nla_nest_start(msg, RTA_METRICS)))
1416                         goto nla_put_failure;
1417
1418                 nla_put_u32(msg, RTAX_MTU, route->mtu);
1419
1420                 nla_nest_end(msg, metrics);
1421         }
1422
1423         return system_rtnl_call(msg);
1424
1425 nla_put_failure:
1426         nlmsg_free(msg);
1427         return -ENOMEM;
1428 }
1429
1430 int system_add_route(struct device *dev, struct device_route *route)
1431 {
1432         return system_rt(dev, route, RTM_NEWROUTE);
1433 }
1434
1435 int system_del_route(struct device *dev, struct device_route *route)
1436 {
1437         return system_rt(dev, route, RTM_DELROUTE);
1438 }
1439
1440 int system_flush_routes(void)
1441 {
1442         const char *names[] = {
1443                 "/proc/sys/net/ipv4/route/flush",
1444                 "/proc/sys/net/ipv6/route/flush"
1445         };
1446         int fd, i;
1447
1448         for (i = 0; i < ARRAY_SIZE(names); i++) {
1449                 fd = open(names[i], O_WRONLY);
1450                 if (fd < 0)
1451                         continue;
1452
1453                 if (write(fd, "-1", 2)) {}
1454                 close(fd);
1455         }
1456         return 0;
1457 }
1458
1459 bool system_resolve_rt_type(const char *type, unsigned int *id)
1460 {
1461         return system_rtn_aton(type, id);
1462 }
1463
1464 bool system_resolve_rt_table(const char *name, unsigned int *id)
1465 {
1466         FILE *f;
1467         char *e, buf[128];
1468         unsigned int n, table = RT_TABLE_UNSPEC;
1469
1470         /* first try to parse table as number */
1471         if ((n = strtoul(name, &e, 0)) > 0 && !*e)
1472                 table = n;
1473
1474         /* handle well known aliases */
1475         else if (!strcmp(name, "default"))
1476                 table = RT_TABLE_DEFAULT;
1477         else if (!strcmp(name, "main"))
1478                 table = RT_TABLE_MAIN;
1479         else if (!strcmp(name, "local"))
1480                 table = RT_TABLE_LOCAL;
1481
1482         /* try to look up name in /etc/iproute2/rt_tables */
1483         else if ((f = fopen("/etc/iproute2/rt_tables", "r")) != NULL)
1484         {
1485                 while (fgets(buf, sizeof(buf) - 1, f) != NULL)
1486                 {
1487                         if ((e = strtok(buf, " \t\n")) == NULL || *e == '#')
1488                                 continue;
1489
1490                         n = strtoul(e, NULL, 10);
1491                         e = strtok(NULL, " \t\n");
1492
1493                         if (e && !strcmp(e, name))
1494                         {
1495                                 table = n;
1496                                 break;
1497                         }
1498                 }
1499
1500                 fclose(f);
1501         }
1502
1503         if (table == RT_TABLE_UNSPEC)
1504                 return false;
1505
1506         *id = table;
1507         return true;
1508 }
1509
1510 bool system_is_default_rt_table(unsigned int id)
1511 {
1512         return (id == RT_TABLE_MAIN);
1513 }
1514
1515 static int system_iprule(struct iprule *rule, int cmd)
1516 {
1517         int alen = ((rule->flags & IPRULE_FAMILY) == IPRULE_INET4) ? 4 : 16;
1518
1519         struct nl_msg *msg;
1520         struct rtmsg rtm = {
1521                 .rtm_family = (alen == 4) ? AF_INET : AF_INET6,
1522                 .rtm_protocol = RTPROT_STATIC,
1523                 .rtm_scope = RT_SCOPE_UNIVERSE,
1524                 .rtm_table = RT_TABLE_UNSPEC,
1525                 .rtm_type = RTN_UNSPEC,
1526                 .rtm_flags = 0,
1527         };
1528
1529         if (cmd == RTM_NEWRULE) {
1530                 rtm.rtm_type = RTN_UNICAST;
1531                 rtm.rtm_flags |= NLM_F_REPLACE | NLM_F_EXCL;
1532         }
1533
1534         if (rule->invert)
1535                 rtm.rtm_flags |= FIB_RULE_INVERT;
1536
1537         if (rule->flags & IPRULE_SRC)
1538                 rtm.rtm_src_len = rule->src_mask;
1539
1540         if (rule->flags & IPRULE_DEST)
1541                 rtm.rtm_dst_len = rule->dest_mask;
1542
1543         if (rule->flags & IPRULE_TOS)
1544                 rtm.rtm_tos = rule->tos;
1545
1546         if (rule->flags & IPRULE_LOOKUP) {
1547                 if (rule->lookup < 256)
1548                         rtm.rtm_table = rule->lookup;
1549         }
1550
1551         if (rule->flags & IPRULE_ACTION)
1552                 rtm.rtm_type = rule->action;
1553         else if (rule->flags & IPRULE_GOTO)
1554                 rtm.rtm_type = FR_ACT_GOTO;
1555         else if (!(rule->flags & (IPRULE_LOOKUP | IPRULE_ACTION | IPRULE_GOTO)))
1556                 rtm.rtm_type = FR_ACT_NOP;
1557
1558         msg = nlmsg_alloc_simple(cmd, NLM_F_REQUEST);
1559
1560         if (!msg)
1561                 return -1;
1562
1563         nlmsg_append(msg, &rtm, sizeof(rtm), 0);
1564
1565         if (rule->flags & IPRULE_IN)
1566                 nla_put(msg, FRA_IFNAME, strlen(rule->in_dev) + 1, rule->in_dev);
1567
1568         if (rule->flags & IPRULE_OUT)
1569                 nla_put(msg, FRA_OIFNAME, strlen(rule->out_dev) + 1, rule->out_dev);
1570
1571         if (rule->flags & IPRULE_SRC)
1572                 nla_put(msg, FRA_SRC, alen, &rule->src_addr);
1573
1574         if (rule->flags & IPRULE_DEST)
1575                 nla_put(msg, FRA_DST, alen, &rule->dest_addr);
1576
1577         if (rule->flags & IPRULE_PRIORITY)
1578                 nla_put_u32(msg, FRA_PRIORITY, rule->priority);
1579         else if (cmd == RTM_NEWRULE)
1580                 nla_put_u32(msg, FRA_PRIORITY, rule->order);
1581
1582         if (rule->flags & IPRULE_FWMARK)
1583                 nla_put_u32(msg, FRA_FWMARK, rule->fwmark);
1584
1585         if (rule->flags & IPRULE_FWMASK)
1586                 nla_put_u32(msg, FRA_FWMASK, rule->fwmask);
1587
1588         if (rule->flags & IPRULE_LOOKUP) {
1589                 if (rule->lookup >= 256)
1590                         nla_put_u32(msg, FRA_TABLE, rule->lookup);
1591         }
1592
1593         if (rule->flags & IPRULE_GOTO)
1594                 nla_put_u32(msg, FRA_GOTO, rule->gotoid);
1595
1596         return system_rtnl_call(msg);
1597 }
1598
1599 int system_add_iprule(struct iprule *rule)
1600 {
1601         return system_iprule(rule, RTM_NEWRULE);
1602 }
1603
1604 int system_del_iprule(struct iprule *rule)
1605 {
1606         return system_iprule(rule, RTM_DELRULE);
1607 }
1608
1609 int system_flush_iprules(void)
1610 {
1611         int rv = 0;
1612         struct iprule rule;
1613
1614         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET);
1615         system_if_clear_entries(NULL, RTM_GETRULE, AF_INET6);
1616
1617         memset(&rule, 0, sizeof(rule));
1618
1619
1620         rule.flags = IPRULE_INET4 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1621
1622         rule.priority = 0;
1623         rule.lookup = RT_TABLE_LOCAL;
1624         rv |= system_iprule(&rule, RTM_NEWRULE);
1625
1626         rule.priority = 32766;
1627         rule.lookup = RT_TABLE_MAIN;
1628         rv |= system_iprule(&rule, RTM_NEWRULE);
1629
1630         rule.priority = 32767;
1631         rule.lookup = RT_TABLE_DEFAULT;
1632         rv |= system_iprule(&rule, RTM_NEWRULE);
1633
1634
1635         rule.flags = IPRULE_INET6 | IPRULE_PRIORITY | IPRULE_LOOKUP;
1636
1637         rule.priority = 0;
1638         rule.lookup = RT_TABLE_LOCAL;
1639         rv |= system_iprule(&rule, RTM_NEWRULE);
1640
1641         rule.priority = 32766;
1642         rule.lookup = RT_TABLE_MAIN;
1643         rv |= system_iprule(&rule, RTM_NEWRULE);
1644
1645         return rv;
1646 }
1647
1648 bool system_resolve_iprule_action(const char *action, unsigned int *id)
1649 {
1650         return system_rtn_aton(action, id);
1651 }
1652
1653 time_t system_get_rtime(void)
1654 {
1655         struct timespec ts;
1656         struct timeval tv;
1657
1658         if (syscall(__NR_clock_gettime, CLOCK_MONOTONIC, &ts) == 0)
1659                 return ts.tv_sec;
1660
1661         if (gettimeofday(&tv, NULL) == 0)
1662                 return tv.tv_sec;
1663
1664         return 0;
1665 }
1666
1667 #ifndef IP_DF
1668 #define IP_DF       0x4000
1669 #endif
1670
1671 static int tunnel_ioctl(const char *name, int cmd, void *p)
1672 {
1673         struct ifreq ifr;
1674
1675         memset(&ifr, 0, sizeof(ifr));
1676         strncpy(ifr.ifr_name, name, sizeof(ifr.ifr_name));
1677         ifr.ifr_ifru.ifru_data = p;
1678         return ioctl(sock_ioctl, cmd, &ifr);
1679 }
1680
1681 #ifdef IFLA_IPTUN_MAX
1682 #define IP6_FLOWINFO_TCLASS     htonl(0x0FF00000)
1683 static int system_add_gre_tunnel(const char *name, const char *kind,
1684                                  const unsigned int link, struct blob_attr **tb, bool v6)
1685 {
1686         struct nl_msg *nlm;
1687         struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC, };
1688         struct blob_attr *cur;
1689         uint32_t ikey = 0, okey = 0, flags = 0, flowinfo = 0;
1690         uint16_t iflags = 0, oflags = 0;
1691         uint8_t tos = 0;
1692         int ret = 0, ttl = 64;
1693
1694         nlm = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
1695         if (!nlm)
1696                 return -1;
1697
1698         nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
1699         nla_put_string(nlm, IFLA_IFNAME, name);
1700
1701         struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
1702         if (!linkinfo) {
1703                 ret = -ENOMEM;
1704                 goto failure;
1705         }
1706
1707         nla_put_string(nlm, IFLA_INFO_KIND, kind);
1708         struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
1709         if (!infodata) {
1710                 ret = -ENOMEM;
1711                 goto failure;
1712         }
1713
1714         if (link)
1715                 nla_put_u32(nlm, IFLA_GRE_LINK, link);
1716
1717         if ((cur = tb[TUNNEL_ATTR_TTL]))
1718                 ttl = blobmsg_get_u32(cur);
1719
1720         nla_put_u8(nlm, IFLA_GRE_TTL, ttl);
1721
1722         if ((cur = tb[TUNNEL_ATTR_TOS])) {
1723                 char *str = blobmsg_get_string(cur);
1724                 if (strcmp(str, "inherit")) {
1725                         unsigned uval;
1726
1727                         if (!system_tos_aton(str, &uval)) {
1728                                 ret = -EINVAL;
1729                                 goto failure;
1730                         }
1731
1732                         if (v6)
1733                                 flowinfo |= htonl(uval << 20) & IP6_FLOWINFO_TCLASS;
1734                         else
1735                                 tos = uval;
1736                 } else {
1737                         if (v6)
1738                                 flags |= IP6_TNL_F_USE_ORIG_TCLASS;
1739                         else
1740                                 tos = 1;
1741                 }
1742         }
1743
1744         if ((cur = tb[TUNNEL_ATTR_INFO]) && (blobmsg_type(cur) == BLOBMSG_TYPE_STRING)) {
1745                 uint8_t icsum, ocsum, iseqno, oseqno;
1746                 if (sscanf(blobmsg_get_string(cur), "%u,%u,%hhu,%hhu,%hhu,%hhu",
1747                         &ikey, &okey, &icsum, &ocsum, &iseqno, &oseqno) < 6) {
1748                         ret = -EINVAL;
1749                         goto failure;
1750                 }
1751
1752                 if (ikey)
1753                         iflags |= GRE_KEY;
1754
1755                 if (okey)
1756                         oflags |= GRE_KEY;
1757
1758                 if (icsum)
1759                         iflags |= GRE_CSUM;
1760
1761                 if (ocsum)
1762                         oflags |= GRE_CSUM;
1763
1764                 if (iseqno)
1765                         iflags |= GRE_SEQ;
1766
1767                 if (oseqno)
1768                         oflags |= GRE_SEQ;
1769         }
1770
1771         if (v6) {
1772                 struct in6_addr in6buf;
1773                 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1774                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1775                                 ret = -EINVAL;
1776                                 goto failure;
1777                         }
1778                         nla_put(nlm, IFLA_GRE_LOCAL, sizeof(in6buf), &in6buf);
1779                 }
1780
1781                 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1782                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
1783                                 ret = -EINVAL;
1784                                 goto failure;
1785                         }
1786                         nla_put(nlm, IFLA_GRE_REMOTE, sizeof(in6buf), &in6buf);
1787                 }
1788                 nla_put_u8(nlm, IFLA_GRE_ENCAP_LIMIT, 4);
1789
1790                 if (flowinfo)
1791                         nla_put_u32(nlm, IFLA_GRE_FLOWINFO, flowinfo);
1792
1793                 if (flags)
1794                         nla_put_u32(nlm, IFLA_GRE_FLAGS, flags);
1795         } else {
1796                 struct in_addr inbuf;
1797                 bool set_df = true;
1798
1799                 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
1800                         if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
1801                                 ret = -EINVAL;
1802                                 goto failure;
1803                         }
1804                         nla_put(nlm, IFLA_GRE_LOCAL, sizeof(inbuf), &inbuf);
1805                 }
1806
1807                 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
1808                         if (inet_pton(AF_INET, blobmsg_data(cur), &inbuf) < 1) {
1809                                 ret = -EINVAL;
1810                                 goto failure;
1811                         }
1812                         nla_put(nlm, IFLA_GRE_REMOTE, sizeof(inbuf), &inbuf);
1813
1814                         if (IN_MULTICAST(ntohl(inbuf.s_addr))) {
1815                                 if (!okey) {
1816                                         okey = inbuf.s_addr;
1817                                         oflags |= GRE_KEY;
1818                                 }
1819
1820                                 if (!ikey) {
1821                                         ikey = inbuf.s_addr;
1822                                         iflags |= GRE_KEY;
1823                                 }
1824                         }
1825                 }
1826
1827                 if ((cur = tb[TUNNEL_ATTR_DF]))
1828                         set_df = blobmsg_get_bool(cur);
1829
1830                 /* ttl !=0 and nopmtudisc are incompatible */
1831                 if (ttl && !set_df) {
1832                         ret = -EINVAL;
1833                         goto failure;
1834                 }
1835
1836                 nla_put_u8(nlm, IFLA_GRE_PMTUDISC, set_df ? 1 : 0);
1837
1838                 nla_put_u8(nlm, IFLA_GRE_TOS, tos);
1839         }
1840
1841         if (oflags)
1842                 nla_put_u16(nlm, IFLA_GRE_OFLAGS, oflags);
1843
1844         if (iflags)
1845                 nla_put_u16(nlm, IFLA_GRE_IFLAGS, iflags);
1846
1847         if (okey)
1848                 nla_put_u32(nlm, IFLA_GRE_OKEY, okey);
1849
1850         if (ikey)
1851                 nla_put_u32(nlm, IFLA_GRE_IKEY, ikey);
1852
1853         nla_nest_end(nlm, infodata);
1854         nla_nest_end(nlm, linkinfo);
1855
1856         return system_rtnl_call(nlm);
1857
1858 failure:
1859         nlmsg_free(nlm);
1860         return ret;
1861 }
1862 #endif
1863
1864 static int system_add_proto_tunnel(const char *name, const uint8_t proto, const unsigned int link, struct blob_attr **tb)
1865 {
1866         struct blob_attr *cur;
1867         bool set_df = true;
1868         struct ip_tunnel_parm p  = {
1869                 .link = link,
1870                 .iph = {
1871                         .version = 4,
1872                         .ihl = 5,
1873                         .protocol = proto,
1874                 }
1875         };
1876
1877         if ((cur = tb[TUNNEL_ATTR_LOCAL]) &&
1878                         inet_pton(AF_INET, blobmsg_data(cur), &p.iph.saddr) < 1)
1879                 return -EINVAL;
1880
1881         if ((cur = tb[TUNNEL_ATTR_REMOTE]) &&
1882                         inet_pton(AF_INET, blobmsg_data(cur), &p.iph.daddr) < 1)
1883                 return -EINVAL;
1884
1885         if ((cur = tb[TUNNEL_ATTR_DF]))
1886                 set_df = blobmsg_get_bool(cur);
1887
1888         if ((cur = tb[TUNNEL_ATTR_TTL]))
1889                 p.iph.ttl = blobmsg_get_u32(cur);
1890
1891         if ((cur = tb[TUNNEL_ATTR_TOS])) {
1892                 char *str = blobmsg_get_string(cur);
1893                 if (strcmp(str, "inherit")) {
1894                         unsigned uval;
1895
1896                         if (!system_tos_aton(str, &uval))
1897                                 return -EINVAL;
1898
1899                         p.iph.tos = uval;
1900                 } else
1901                         p.iph.tos = 1;
1902         }
1903
1904         p.iph.frag_off = set_df ? htons(IP_DF) : 0;
1905         /* ttl !=0 and nopmtudisc are incompatible */
1906         if (p.iph.ttl && p.iph.frag_off == 0)
1907                 return -EINVAL;
1908
1909         strncpy(p.name, name, sizeof(p.name));
1910
1911         switch (p.iph.protocol) {
1912         case IPPROTO_IPIP:
1913                 return tunnel_ioctl("tunl0", SIOCADDTUNNEL, &p);
1914         case IPPROTO_IPV6:
1915                 return tunnel_ioctl("sit0", SIOCADDTUNNEL, &p);
1916         default:
1917                 break;
1918         }
1919         return -1;
1920 }
1921
1922 static int __system_del_ip_tunnel(const char *name, struct blob_attr **tb)
1923 {
1924         struct blob_attr *cur;
1925         const char *str;
1926
1927         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
1928                 return -EINVAL;
1929         str = blobmsg_data(cur);
1930
1931         if (!strcmp(str, "greip") || !strcmp(str, "gretapip") ||
1932             !strcmp(str, "greip6") || !strcmp(str, "gretapip6"))
1933                 return system_link_del(name);
1934         else
1935                 return tunnel_ioctl(name, SIOCDELTUNNEL, NULL);
1936 }
1937
1938 int system_del_ip_tunnel(const char *name, struct blob_attr *attr)
1939 {
1940         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
1941
1942         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
1943                 blob_data(attr), blob_len(attr));
1944
1945         return __system_del_ip_tunnel(name, tb);
1946 }
1947
1948 int system_update_ipv6_mtu(struct device *dev, int mtu)
1949 {
1950         int ret = -1;
1951         char buf[64];
1952         snprintf(buf, sizeof(buf), "/proc/sys/net/ipv6/conf/%s/mtu",
1953                         dev->ifname);
1954
1955         int fd = open(buf, O_RDWR);
1956         ssize_t len = read(fd, buf, sizeof(buf) - 1);
1957         if (len < 0)
1958                 goto out;
1959
1960         buf[len] = 0;
1961         ret = atoi(buf);
1962
1963         if (!mtu || ret <= mtu)
1964                 goto out;
1965
1966         lseek(fd, 0, SEEK_SET);
1967         if (write(fd, buf, snprintf(buf, sizeof(buf), "%i", mtu)) <= 0)
1968                 ret = -1;
1969
1970 out:
1971         close(fd);
1972         return ret;
1973 }
1974
1975 int system_add_ip_tunnel(const char *name, struct blob_attr *attr)
1976 {
1977         struct blob_attr *tb[__TUNNEL_ATTR_MAX];
1978         struct blob_attr *cur;
1979         const char *str;
1980
1981         blobmsg_parse(tunnel_attr_list.params, __TUNNEL_ATTR_MAX, tb,
1982                 blob_data(attr), blob_len(attr));
1983
1984         __system_del_ip_tunnel(name, tb);
1985
1986         if (!(cur = tb[TUNNEL_ATTR_TYPE]))
1987                 return -EINVAL;
1988         str = blobmsg_data(cur);
1989
1990         unsigned int ttl = 0;
1991         if ((cur = tb[TUNNEL_ATTR_TTL])) {
1992                 ttl = blobmsg_get_u32(cur);
1993                 if (ttl > 255)
1994                         return -EINVAL;
1995         }
1996
1997         unsigned int link = 0;
1998         if ((cur = tb[TUNNEL_ATTR_LINK])) {
1999                 struct interface *iface = vlist_find(&interfaces, blobmsg_data(cur), iface, node);
2000                 if (!iface)
2001                         return -EINVAL;
2002
2003                 if (iface->l3_dev.dev)
2004                         link = iface->l3_dev.dev->ifindex;
2005         }
2006
2007         if (!strcmp(str, "sit")) {
2008                 if (system_add_proto_tunnel(name, IPPROTO_IPV6, link, tb) < 0)
2009                         return -1;
2010
2011 #ifdef SIOCADD6RD
2012                 if ((cur = tb[TUNNEL_ATTR_6RD_PREFIX])) {
2013                         unsigned int mask;
2014                         struct ip_tunnel_6rd p6;
2015
2016                         memset(&p6, 0, sizeof(p6));
2017
2018                         if (!parse_ip_and_netmask(AF_INET6, blobmsg_data(cur),
2019                                                 &p6.prefix, &mask) || mask > 128)
2020                                 return -EINVAL;
2021                         p6.prefixlen = mask;
2022
2023                         if ((cur = tb[TUNNEL_ATTR_6RD_RELAY_PREFIX])) {
2024                                 if (!parse_ip_and_netmask(AF_INET, blobmsg_data(cur),
2025                                                         &p6.relay_prefix, &mask) || mask > 32)
2026                                         return -EINVAL;
2027                                 p6.relay_prefixlen = mask;
2028                         }
2029
2030                         if (tunnel_ioctl(name, SIOCADD6RD, &p6) < 0) {
2031                                 __system_del_ip_tunnel(name, tb);
2032                                 return -1;
2033                         }
2034                 }
2035 #endif
2036 #ifdef IFLA_IPTUN_MAX
2037         } else if (!strcmp(str, "ipip6")) {
2038                 struct nl_msg *nlm = nlmsg_alloc_simple(RTM_NEWLINK,
2039                                 NLM_F_REQUEST | NLM_F_REPLACE | NLM_F_CREATE);
2040                 struct ifinfomsg ifi = { .ifi_family = AF_UNSPEC };
2041                 int ret = 0;
2042
2043                 if (!nlm)
2044                         return -1;
2045
2046                 nlmsg_append(nlm, &ifi, sizeof(ifi), 0);
2047                 nla_put_string(nlm, IFLA_IFNAME, name);
2048
2049                 if (link)
2050                         nla_put_u32(nlm, IFLA_LINK, link);
2051
2052                 struct nlattr *linkinfo = nla_nest_start(nlm, IFLA_LINKINFO);
2053                 if (!linkinfo) {
2054                         ret = -ENOMEM;
2055                         goto failure;
2056                 }
2057                 nla_put_string(nlm, IFLA_INFO_KIND, "ip6tnl");
2058                 struct nlattr *infodata = nla_nest_start(nlm, IFLA_INFO_DATA);
2059                 if (!infodata) {
2060                         ret = -ENOMEM;
2061                         goto failure;
2062                 }
2063
2064                 if (link)
2065                         nla_put_u32(nlm, IFLA_IPTUN_LINK, link);
2066
2067                 nla_put_u8(nlm, IFLA_IPTUN_PROTO, IPPROTO_IPIP);
2068                 nla_put_u8(nlm, IFLA_IPTUN_TTL, (ttl) ? ttl : 64);
2069                 nla_put_u8(nlm, IFLA_IPTUN_ENCAP_LIMIT, 4);
2070
2071                 struct in6_addr in6buf;
2072                 if ((cur = tb[TUNNEL_ATTR_LOCAL])) {
2073                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2074                                 ret = -EINVAL;
2075                                 goto failure;
2076                         }
2077                         nla_put(nlm, IFLA_IPTUN_LOCAL, sizeof(in6buf), &in6buf);
2078                 }
2079
2080                 if ((cur = tb[TUNNEL_ATTR_REMOTE])) {
2081                         if (inet_pton(AF_INET6, blobmsg_data(cur), &in6buf) < 1) {
2082                                 ret = -EINVAL;
2083                                 goto failure;
2084                         }
2085                         nla_put(nlm, IFLA_IPTUN_REMOTE, sizeof(in6buf), &in6buf);
2086                 }
2087
2088 #ifdef IFLA_IPTUN_FMR_MAX
2089                 if ((cur = tb[TUNNEL_ATTR_FMRS])) {
2090                         struct nlattr *fmrs = nla_nest_start(nlm, IFLA_IPTUN_FMRS);
2091
2092                         struct blob_attr *fmr;
2093                         unsigned rem, fmrcnt = 0;
2094                         blobmsg_for_each_attr(fmr, cur, rem) {
2095                                 if (blobmsg_type(fmr) != BLOBMSG_TYPE_STRING)
2096                                         continue;
2097
2098                                 unsigned ip4len, ip6len, ealen, offset = 6;
2099                                 char ip6buf[48];
2100                                 char ip4buf[16];
2101
2102                                 if (sscanf(blobmsg_get_string(fmr), "%47[^/]/%u,%15[^/]/%u,%u,%u",
2103                                                 ip6buf, &ip6len, ip4buf, &ip4len, &ealen, &offset) < 5) {
2104                                         ret = -EINVAL;
2105                                         goto failure;
2106                                 }
2107
2108                                 struct in6_addr ip6prefix;
2109                                 struct in_addr ip4prefix;
2110                                 if (inet_pton(AF_INET6, ip6buf, &ip6prefix) != 1 ||
2111                                                 inet_pton(AF_INET, ip4buf, &ip4prefix) != 1) {
2112                                         ret = -EINVAL;
2113                                         goto failure;
2114                                 }
2115
2116                                 struct nlattr *rule = nla_nest_start(nlm, ++fmrcnt);
2117
2118                                 nla_put(nlm, IFLA_IPTUN_FMR_IP6_PREFIX, sizeof(ip6prefix), &ip6prefix);
2119                                 nla_put(nlm, IFLA_IPTUN_FMR_IP4_PREFIX, sizeof(ip4prefix), &ip4prefix);
2120                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP6_PREFIX_LEN, ip6len);
2121                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_IP4_PREFIX_LEN, ip4len);
2122                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_EA_LEN, ealen);
2123                                 nla_put_u8(nlm, IFLA_IPTUN_FMR_OFFSET, offset);
2124
2125                                 nla_nest_end(nlm, rule);
2126                         }
2127
2128                         nla_nest_end(nlm, fmrs);
2129                 }
2130 #endif
2131
2132                 nla_nest_end(nlm, infodata);
2133                 nla_nest_end(nlm, linkinfo);
2134
2135                 return system_rtnl_call(nlm);
2136 failure:
2137                 nlmsg_free(nlm);
2138                 return ret;
2139         } else if (!strcmp(str, "greip")) {
2140                 return system_add_gre_tunnel(name, "gre", link, tb, false);
2141         } else if (!strcmp(str, "gretapip"))  {
2142                 return system_add_gre_tunnel(name, "gretap", link, tb, false);
2143         } else if (!strcmp(str, "greip6")) {
2144                 return system_add_gre_tunnel(name, "ip6gre", link, tb, true);
2145         } else if (!strcmp(str, "gretapip6")) {
2146                 return system_add_gre_tunnel(name, "ip6gretap", link, tb, true);
2147 #endif
2148         } else if (!strcmp(str, "ipip")) {
2149                 return system_add_proto_tunnel(name, IPPROTO_IPIP, link, tb);
2150         }
2151         else
2152                 return -EINVAL;
2153
2154         return 0;
2155 }